AMD GPU14 Tech Event Sept 25 - AMD Hawiian Islands

[cmdrdredd]

You said valve is sort of partnered with nvidia. I fail to see the direct connection. Amd also talked up Linux in the midst of the valve announcements.

Nvidia has more to gain there...they don't have their GPUs in the PS4 or XB1.

 

[Xarick]

John Carmac weighs in?

https://twitter.com/ID_AA_Carmack/status/383051142755663872

 

[geokilla]

So I haven't really followed the announcement and the GPU world lately. In short, I guess this was a very successful announcement?

 

[VulgarDisplay]

Nvidia has more to gain there...they don't have their GPUs in the PS4 or XB1.

Amd has far more to gain than nvidia. They have been hemorrhaging money and market share because they were unable to compete in the markets shaped by Intel and nvidia. Since they could not beat them at their game amd has made its smartest move in years and cornered the market on gaming.

 

[tigersty1e]

how the 7970/50 benches on bf4 compared to nvidia's card will be very important.

nvidia has the edge over AMD in bf3.

but if amd can eek out a win on this game, then it will be a good early indicator of mantle's future.

 

[BallaTheFeared]

John Carmac weighs in?

https://twitter.com/ID_AA_Carmack/status/383051142755663872

He's taking questions, someone with twitter ask him if it's likely Nvidia would do similar, and how much fragmentation this might cause the industry.

Also ask if it's possible for M$ to do the same with DX, because if EA and AMD can do, it's probably not rocket science, or is it just the product of a monopoly on consoles leaking into the elitist ecosystem of PCs?

how the 7970/50 benches on bf4 compared to nvidia's card will be very important.

nvidia has the edge over AMD in bf3.

but if amd can eek out a win on this game, then it will be a good early indicator of mantle's future.

I wonder how much access Nvidia has to the client, or if a large majority of PC users will suffer from a month or more of poor performance due to this alliance between EA and AMD.



Personally I think it's more related to CPUs, specifically AMD's slow Desktop and now Console eight core offerings. It would seem to me this is more to address that.

It will be interesting to see if AMD users can force DX11 with non bottlenecking CPUs to see how much actual benefit the API provides outside of non cpu limited situations.

 

[cmdrdredd]

Amd has far more to gain than nvidia. They have been hemorrhaging money and market share because they were unable to compete in the markets shaped by Intel and nvidia. Since they could not beat them at their game amd has made its smartest move in years and cornered the market on gaming.

You misread...I'm not talking money.

John Carmac weighs in?

https://twitter.com/ID_AA_Carmack/status/383051142755663872

Based on what he said...if it works in Valve's favor the steambox may get buried. MS and Sony have too much money and industry recognition in the gaming industry. I don't know how well it would work on linux though.

 

[VulgarDisplay]

You misread...I'm not talking money.

You were talking market share which is essentially the same thing.

 

[cmdrdredd]

You were talking market share which is essentially the same thing.

nope

I'm talking about getting your name out there. Not selling product.

 

[blackened23]

The amount of FUD being promoted in this thread is unbelievable

First of all, CUDA is not an API like "Mantle". CUDA is a "parallel computing platform and programming model" that makes it easier to use an NVIDIA GPU for general purpose computing and for extracting more parallelism (http://blogs.nvidia.com/blog/2012/09/10/what-is-cuda-2/). At the time that CUDA was created, there was nothing else available to effectively do this.

Second of all, "Mantle" as an API is destined to fail in the marketplace for a variety of reasons. One is that the majority of PC gamers and PC Game developers (including John Carmack and Tim Sweeney) prefer NVIDIA GPU's + Intel CPU's compared to AMD GPU's/CPU's. Two is that open platform gaming (which includes Windows PC, Android PC, SteamOS, etc) is growing. Three is that ultra mobile platform gaming is growing, and NVIDIA/Intel/Qualcomm are way ahead of AMD in that area. For example, the tremendous efficiency gains that NVIDIA is seeing in the ultra mobile GPU space starting with the Kepler.M GPU in "Logan" will filter through their entire lineup in the near future with Maxwell and beyond.

Just to piggyback on this comment, it should be noted that Carmack, Sweeney and Timothy Lottes (Former NV engineer, now works with Epic games) have spoken disgust with regards to the API situation on PCs and how much performance it "leaves on the table", so to speak. Carmack in particular has stated that the overhead of DirectX is substantial and with direct hardware access, we would see performance increases in order of 10-15 times the current performance.

That being said, I don't see this happening primarily because it excludes half of the PC playerbase. Developers obviously will want to create games for 100% of the player base, and if nvidia cannot do it while AMD can, that leaves the future uncertain. AMD will _really_ have to do work to convince developers, although I will say - BF4 will sell 70 gabillion copies and that is a good start.

So basically , the future of Mantle depends on whether the performance gains are real (if it *is* true hardware access, 10 times more performance *is* possible) but I suspect that it isn't full direct to metal programming. There has to be a catch here. The future also depends on how well AMD can leverage their AMD GE brand and how developers react to that, and how well the next-gen consoles do (and the ease of porting from consoles to Mantle).

I think the concept behind mantle is awesome. The PC needs this. But the question is how to do this while not being available for over half of the PC playerbase? That would relegate it to Glide status which really had VERY FEW specific titles. I think there were overall 15 glide only titles for the duration of its life.

 

[cmdrdredd]

Just to piggyback on this comment, it should be noted that Carmack, Sweeney and Timothy Lottes (Former NV engineer, now works with Epic games) have spoken disgust with regards to the API situation on PCs and how much performance it "leaves on the table", so to speak. Carmack in particular has stated that the overhead of DirectX is substantial and with direct hardware access, we would see performance increases in order of 10-15 times the current performance.

That being said, I don't see this happening primarily because it excludes half of the PC playerbase. Developers obviously will want to create games for 100% of the player base, and if nvidia cannot do it while AMD can, that leaves the future uncertain. AMD will _really_ have to do work to convince developers, although I will say - BF4 will sell 70 gabillion copies and that is a good start.

So basically , the future of Mantle depends on whether the performance gains are real (if it *is* true hardware access, 10 times more performance *is* possible) but I suspect that it isn't full direct to metal programming. There has to be a catch here. The future also depends on how well AMD can leverage their AMD GE brand and how developers react to that, and how well the next-gen consoles do (and the ease of porting from consoles to Mantle).

I think the concept behind mantle is awesome. The PC needs this. But the question is how to do this while not being available for over half of the PC playerbase? That would relegate it to Glide status which really had VERY FEW specific titles. I think there were overall 15 glide only titles for the duration of its life.

True, but to say one thing in regard to selling millions of copies of BF4. Many of those sales aren't on PC and those that are, what percentage of those consumers are using the proper GPU for Mantle support?

 

[blackened23]

Just to add on to the above post, this is what Timothy Lottes, former software engineer at nVidia and creator of FXAA has stated in regard to direct hardware access:

The real reason to get excited about a PS4 is what Sony as a company does with the OS and system libraries as a platform, and what this enables 1st party studios to do, when they make PS4-only games. If PS4 has a real-time OS, with a libGCM style low level access to the GPU, then the PS4 1st party games will be years ahead of the PC simply because it opens up what is possible on the GPU. Note this won't happen right away on launch, but once developers tool up for the platform, this will be the case. As a PC guy who knows hardware to the metal, I spend most of my days in frustration knowing damn well what I could do with the hardware, but what I cannot do because Microsoft and IHVs wont provide low-level GPU access in PC APIs. One simple example, drawcalls on PC have easily 10x to 100x the overhead of a console with a libGCM style API.

I think anyone dismissing the concept of direct hardware programming are just in the wrong. There are very real gains that can be done with libGCM style programming and those performance gains are unreal. But the real question is, how to do this on the PC? How can AMD leverage their brand? Basically this style would relegate the PC to "closed wall" type boxes where you MUST have a specific brand of CPU or GPU to take advantage of it. Many consumers will obviously have issues with this, and most developers want to sell everyone a game, not just a portion of the PC audience.

It's going to be tough. The PC fundamentally is and always HAS been an open architecture, and this type of programming flies counter to that despite any performance gains that are possible.

 

[BallaTheFeared]

Just to add on to the above post, this is what Timothy Lottes, former software engineer at nVidia.

Can't imagine why he's a "former" employee at nVidia ()

 

[cmdrdredd]

Can't imagine why he's a "former" employee at nVidia ()

There's lots of former employees of successful companies who left for their own reasons.

 

[blackened23]

Well you can speculate all day long on that and i'm sure I know what your response will be , but Mr. Lottes stated he left on his own volition because he wanted to work directly in the games industry - he resigned from NV to work at Epic games, where he still is today. (By the way, we have Mr. Lottes to thank for FXAA - he was the sole creator of it). He obviously knows his stuff (as does John Carmack, who mirrored what Mr. Lottes said about direct hardware access) and neither of these guys are idiots. They're well respected in the industry and have the credentials to back up anything they say.

My only point here is that dismissing the concept of direct hardware programming via LibGCN is idiotic. It is a sound principle that can open up what the PC can do (and increase performance dramatically), but it may end up failing because of the PC has always been an open architecture. That doesn't appear to be changing anytime soon as developers don't want to develop for only a small percentage of the user base, they want all of the user base to buy their games. So it will be interesting to see what happens - bf4 will be a great first title to show it off, I suppose. But I really think this will be relegated to Glide status despite being a sound new programming principle; IIRC there weren't an awful lots of glide only games. 15 or so? If I remember correctly.

This is aside from the fact that we don't have full details on Mantle yet. We don't know that it is fully direct hardware access. Maybe it isn't. That's we need more information and more documentation.

 

[AnandThenMan]

How can anyone question the merits of close to metal programming? Look what current consoles eek out of anemic hardware, it's really unbelievable.

 

[blackened23]

How can anyone question the merits of close to metal programming? Look what current consoles eek out of anemic hardware, it's really unbelievable.

Like I said it is a sound concept. As much as I welcome the change - I'd have to say that it may be doomed to fail because the PC is not a closed architecture and never will be. If it does take off, I can't see it being anything more than glide status with 2-3 titles per year taking advantage of it. I really don't think this will be a game changer for AMD, although I do welcome the fact that they worked on this. The PC needs this (direct hardware access). But everyone needs it, not just AMD. I just had the craziest idea - maybe NV and AMD could be friends for a day and cross license physx and Mantle? Nahhh, that will never happen.

 

[BallaTheFeared]

I think the GPU enhancements are PR spin and the real benefit is to draw call reduction and higher threaded cpu utilization (Up to eight cores).

AMD claims Mantle enables nine times more draw calls per second than other APIs, which is a huge increase in performance.

AMD hasn't claimed 10% more GPU performance, or anything similar. The only thing they've discussed was cpu core usage and draw call reduction. Which was specifically targeted at reducing CPU overhead and perhaps some memory.

The 10x to 100x is a simply a draw call limitation and AMD is getting nine times more draw calls from the same CPUs with this new API which is a far cry for the 10 to 100 times others have stated. Which again has nothing to do directly with GPU performance increases, unless CPU limited.

 

[RussianSensation]

Excellent post jvroig. :thumbsup:

Seconded! :thumbsup:

Well we went from rumours of a new card that could be 2X as fast as Titan, to on par with Titan, to today's announcement.......

No one with an IQ of over 95 would believe that AMD could produce a 28nm card 2x faster than Titan. Most reasonable rumors pointed to performance between a reference 780 and Titan.

Again, just like not many people would switch from 6800U to X850XT PE, 4890 to 285, GTX680 to 7970GE or from a 680 to a 770, R9 290X is not meant for people to sell their 780/Titan because the performance increase is not enough. It's there to make sure AMD has a competitive high-end offering for the next 9-10 months to compete against NV's finest and whatever else NV comes up with in the near future. Gamers who are going to be upgrading for BF4 for example may then consider 780 vs. R9 290X. That's the point.

In fact, for anyone who bought GTX680 and 7970, the 2nd half of 28nm generation of GPUs is not very exciting either. If you want excitement and an increase of 75-100% from today's cards, you have to wait for 20nm Maxwell.

AMD >>better<< not use GCN AGAIN at 20nm.

If they aren't using GCN for their 20nm GPUs, they won't get any benefit from this stuff for more than literally half a year give or take.

Trying to fight a new architecture like Maxwell with GCN AGAIN is just madness.

What? They used VLIW from 2006 to 2010, or nearly 5 years. GCN is not about to be replaced. It will continue to evolve. Also, look at NV, they used Fermi for nearly 4 years now because Kepler is fundamentally just a Kepler architecture with revised shader clocks and other optimizations. NV also uses its GPU architectures for 4-5 years. Maxwell should be their brand new compute-based architecture built from the ground-up according to rumors. NV did well with Kepler despite its roots tracing back to at least 2009.

 

[brandonmatic]

AMD hasn't claimed 10% more GPU performance, or anything similar. The only thing they've discussed was cpu core usage and draw call reduction. Which was specifically targeted at reducing CPU overhead and perhaps some memory.

Here's the exact text from the relevant Mantle slide (123):

1. Enables 9X more draw calls per second than other APIs by reducing CPU overhead
2. Enables higher graphics performance with direct access to all GPU features
3. New rendering techniques
4. Leverage optimization work from next-gen game consoles to PCs

https://amd.app.box.com/GPU14publicpreso

 

[Carfax83]

Regarding what Lottes said, is there a specific reason why Microsoft doesn't want to provide lower level access to GPU hardware via their API?

Or is it a case of they just haven't evolved to that stage yet?

 

[Jodell88]

Nvidia has committed to help with linux driver support. http://www.anandtech.com/show/7367/nvidia-pledges-help-on-open-source-linux-gpu-driver-development

This announcement just happens to fall the day after SteamOS reveal. Right now there's nothing definite, but...

You're reading way too much into that announcement. That is only for a open source driver that Nvidia has nothing to do with.

 

[HurleyBird]

AMD hasn't claimed 10% more GPU performance, or anything similar. The only thing they've discussed was cpu core usage and draw call reduction. Which was specifically targeted at reducing CPU overhead and perhaps some memory.

"Really highly optimized GPU usage."

"Enables developers to explore advanced rendering techniques not possible now."

"Will give superb GPU performance, better scaling"

 

[blackened23]

Regarding what Lottes said, is there a specific reason why Microsoft doesn't want to provide lower level access to GPU hardware via their API?

Or is it a case of they just haven't evolved to that stage yet?

Because it requires a "closed box" with various types of specific hardware. MS obviously doesn't want that to happen - they want a general API that serves any and all needs which is precisely what DirectX does. You can use any GPU, any CPU with directX. Direct to hardware programming is not hardware agnostic while DX is.

DX definitely has shortcomings in terms of potential performance, but MS is trying to serve the greater good? If you believe their company line. That said, we don't know details of Mantle yet. We don't know if it is direct hardware access. They did not provide full documentation on it, so we'll have to wait and see.

 

[BallaTheFeared]

"Really highly optimized GPU usage."

"Enables developers to explore advanced rendering techniques not possible now."

"Will give superb GPU performance, better scaling"

Which means what?

When marketing doesn't spell it out, they're side stepping.